Failure Prediction for Scientific Workflows Using Nature-Inspired Machine Learning Approach

Scientific workflows have gained the emerging attention in sophisti-cated large-scale scientific problem-solving environments.The pay-per-use model of cloud,its scalability and dynamic deployment enables it suited for executing scientific workflow applications.Since the cloud is not a utopian environment,failures are inevitable that may result in experiencingfluctuations in the delivered performance.Though a single task failure occurs in workflow based applications,due to its task dependency nature,the reliability of the overall system will be affected drastically.Hence rather than reactive fault-tolerant approaches,proactive measures are vital in scientific workflows.This work puts forth an attempt to con-centrate on the exploration issue of structuring a nature inspired metaheuristics-Intelligent Water Drops Algorithm(IWDA)combined with an efficient machine learning approach-Support Vector Regression(SVR)for task failure prognostica-tion which facilitates proactive fault-tolerance in the scheduling of scientific workflow applications.The failure prediction models in this study have been implemented through SVR-based machine learning approaches and the precision accuracy of prediction is optimized by IWDA and several performance metrics were evaluated on various benchmark workflows.The experimental results prove that the proposed proactive fault-tolerant approach performs better compared with the other existing techniques.

Iron isotopic analyses of geological reference materials on MC-ICP-MS with instrumental mass bias corrected by three independent methods

Here we report iron(Fe) isotopic data of three pure Fe solution standards(IRMM-014, GSB Fe, and NIST3126a) and five widely used geological reference materials(RMs) from the United States Geological Survey and Geological Survey of Japan obtained on a Neptune Plus multi-collector–inductively coupled plasma–mass spectrometer(MC-ICP-MS) in our laboratory over the past 3 years. The instrumental mass bias was corrected by three independent methods: sample-standard bracketing(SSB),Ni doping + SSB, and ^(57)Fe–^(58)Fe double spike + SSB.Measurements reveal that both the Ni doping and double spike methods helped calibrate short-term fluctuations in mass bias. Collectively, almost all measurements of RMs yielded δ^(56)Fe within ± 0.05 of recommended values,provided that each sample was measured four times on MC-ICP-MS. For the first time, new recommended values for NIST SRM3126a are reported(δ^(56)Fe = 0.363 ± 0.006,2SE, 95% CI; and δ^(57)Fe = 0.534 ± 0.010, 2SE).

A re-assessment of nickel-doping method in iron isotope analysis on rock samples using multi-collector inductively coupled plasma mass spectrometry

Element doping has been proved to be a useful method to correct for the mass bias fractionation when analyzing iron isotope compositions.We present a systematic re-assessment on how the doped nickel may affect the iron isotope analysis in this study by carrying out several experiments.We find three important factors that can affect the analytical results,including the Ni:Fe ratio in the analyte solutions,the match of the Ni:Fe ratio between the unknown sample and standard solutions,and the match of the Fe concentration between the sample and standard solutions.Thus,caution is required when adding Ni to the analyte Fe solutions before analysis.Using our method,theδ56Fe and δ57Fe values of the USGS standards W-2 a,BHVO-2,BCR-2,AGV-2 and GSP-2 are consistent with the recommended literature values,and the long-term(one year) external reproducibility is better than 0.03 and 0.05‰(2 SD) for δ56Fe and δ57Fe,respectively.Therefore,the analytical method established in our laboratory is a method of choice for high quantity Fe isotope data in geological materials.

Development and Validation of a Liquid Chromatography-Tandem Mass Spectrometry Method for Determination of Artemisinin in Rat Plasma

Artemisinin is a potent anti-malarial drug isolated from traditional Chinese medicinal herb, Artemisia annua. The objective of this study was to develop and validate a sensitive and specific LC-MS/MS method for the determination of artemisinin in rat plasma using amlodipine as Internal Standard. The method consist of a simple liquid-liquid extraction with methyl tertiary butyl ether （MTBE） with subsequent evaporation of the supernatant to dryness followed by the analysis of the reconstituted sample by LC-MS/？vIS with a Z-spray atmospheric pressure ionization （API） interface in the positive ion-multiple reaction monitoring mode to monitor precursor--〉product ions of m/z 282.70--〉m/z 209.0 for artemisinin and m/z 408.9--〉m/z 237.0 for amlodipine respectively. The method was linear （0.999） over the concentration range of 7.8-2000 ng/mL in rat plasma. The intra and inter-day accuracy were measured to be within 94-104.2% and precision （CV） were all less than 5%. The extraction recovery means for internal standard and all the artemisinin concentrations used were between 82-85%.

Performance evaluation of fixed-station sampling design for a fishery-independent survey with multiple objectives

Fixed-station sampling design was widely used in fishery-independent surveys because of its characteristics of convenient sampling station setting,but the non-probabilistic(fixed)nature made it more uncertainty of drawing inferences on population.The performance of fixed-station sampling design for multispecies survey has not been evaluated,and we are uncertain if the design could detect the temporal trends of different populations in multispecies fishery-independent survey.In this study,spatial distribution of abundance indices for three species with different spatial distribution patterns including small yellow croaker(Larimichthys polyactis),whitespotted conger(Conger myriaster)and Fang’s blenny(Enedrias fangi)were simulated using ordinary kriging interpolation as the“true”population distribution.The performance of fixed-station sampling design was compared with simple random sampling design by resampling the simulated“true”populations in this simulation study.The results showed that the fixed-station sampling design had the power to detect the seasonal trends of species abundance.The effectiveness of fixed-station sampling design were different in different species distribution patterns.When the species had even distribution,fixed-station sampling design could get high quality abundance data;when the distribution was uneven with heterogeneity or patchiness,fixed-station sampling design tended to underestimate or overestimate the abundance.Evidently,the estimates of abundance index based on the fixedstation sampling design must be calibrated cautiously while applying them for fisheries stock assessment and management.This study suggested that fixed-station sampling design could catch the temporal dynamics of population abundance,but the abundance estimates from the fixed-station sampling design could not be treated as the absolute estimates of populations.

An evaluation of the effects of sample size on estimating length composition of catches from tuna longline fisheries using computer simulations

Length composition analysis can provide insights into the dynamics of a fish population.Accurate quantification of the size structure of a population is critical to understand the status of a fishery and how the population responds to environmental stressors.A scientific observer program is a reliable way to provide such accurate information.However,100%observer coverage is usually impossible for most fisheries because of logistic and financial constraints.Thus,there is a need to evaluate observer program performance,identify suitable sample sizes,and optimize the allocation of observation efforts.The objective of this study is to evaluate the effects of sample size on the quality of length composition data and identify an optimal coverage rate and observation ratio to improve the observation efficiency using an onboard observer data set from China's tuna longline fishery in the western and central Pacific Ocean.We found that the required sample size varies with fish species,indices used to describe length composition,the acceptable accuracy of the estimates,and the allocation methods of sampling effort.Ignoring other information requirements,1000 individuals would be sufficient for most species to reliably quantify length compositions,and a smaller sample size could generate reliable estimates of mean length.A coverage rate of 20%would be sufficient for most species,but a lower coverage rate(5%or 10%)could also be effective to meet with the accuracy and precision requirement in estimating length compositions.A nonrandom effort allocation among fishing baskets within a set could cause the length composition to be overestimated or underestimated for some species.The differences in effective sample sizes among species should be included in the consideration for a rational allocation of observation effort among species when there are different species management priorities.